Fraud control in canephora coffee by NIR and one-class classification method: a study with Amazonian Robustas beans


  • Venancio Ferreira de Moraes Neto
  • Michel Rocha Baqueta
  • Felipe Bachion de Santana
  • Enrique Anastácio Alves
  • Juliana Pallone


Near Infrared Spectroscopy, Chemometrics, Green coffee


The canephora coffee produced in the state of Rondônia - Brazil, the Amazonian Robusta
coffee, obtained a geographical indication of the denomination of origin type, which is
granted to products that have unique characteristics due to the geographical environment,
including natural and human factors. Amazonian Robusta coffees have higher
commercial value, thus the fraudulent practice of labeling low-cost coffees as Amazonian
Robusta coffees needs detection. In this context, coffee authentication is a challenge using
traditional analytical techniques. Thus, the present study aimed to develop a method,
using near infrared spectroscopy (NIR) combined with data driven-soft independent
modeling of class analogies (DD-SIMCA), a one-class classification method, recently
developed for non-destructive authentication of green beans from Amazonian Robustas
coffees. To this end, spectra of Amazonian Robusta coffees (n=114), canephora coffees
from other regions (n=108) and arabica coffees (n=12) were collected without sample
preparation. The samples were directly analyzed by diffuse reflectance in a FT-NIR
system, using a NIRA accessory equipped with a spinning sample module. The scanning
spectra were obtained in triplicate in the spectral range between 10,000 cm-1 and 4,000
cm-1, with a resolution of 4 cm-1 and 16 scans. The samples were separated into training
(70% of the samples) and test (30%) sets by the Kennard-Stone algorithm. The
classification model built based on the full spectra showed 100% correct assignments for
Amazonian Robustas samples and for other coffees, respectively, correctly recognizing
all samples in the training and test sets. In this sense, the combination of NIR and DDSIMCA proved efficient to control the authenticity of the studied coffees. Therefore, the
proposed methodology can be useful for applications in quality control and origin
certification procedures for Amazonian Robustas beans with geographical indication of
the denomination of origin type by direct analysis of the samples, without any type of




Como Citar

Venancio Ferreira de Moraes Neto, Michel Rocha Baqueta, Felipe Bachion de Santana, Enrique Anastácio Alves, & Juliana Pallone. (2022). Fraud control in canephora coffee by NIR and one-class classification method: a study with Amazonian Robustas beans. 1° ongresso e Segurança ualidade os limentos, 1(1). ecuperado de